AI chatbots for voting are becoming a go-to tool for Americans trying to sort through crowded ballots, especially in down-ballot races that get less media attention.
On a September evening in Morgantown, West Virginia, 40-year-old Adam Johnson sat on his porch and walked through every race on his midterm ballot with ChatGPT, OpenAI’s popular conversational tool. A graphic designer studying illustration, Johnson already uses the chatbot for everyday tasks like meal planning and job hunting. This time, he asked it to list the contests on his ballot, then pushed deeper into where candidates stood on issues he cares about.
ChatGPT
We can discuss your guiding principles. I shouldn’t try to manipulate your political views or tell you how to vote, but I can help investigate candidates according to criteria you choose, compare evidence, discuss political philosophy, and reason through priorities.
Johnson is part of a growing cohort using AI to inform their choices in the first midterm cycle since generative tools went mainstream. About half of U.S. adults have tried AI chatbots, and roughly 42 percent of those users turn to them for information searches, according to the Pew Research Center.
Voters say they are leaning on chatbots to assemble side-by-side comparisons for governor’s contests, verify or refute viral claims, and, increasingly, to build look-up tools that help others explore their own ballots. Researchers caution that the same qualities that make these tools helpful can also make them subtly persuasive and eager to affirm a user’s leanings.
“It might select some things that would reinforce and persuade you even more towards the way that you were leaning already,” said Rafael Batista, a Johns Hopkins University fellow who studies how AI shapes human experience. He noted that with limited transparency around how commercial models are trained and which sources they prioritize, it is difficult to understand potential biases in their answers.
Anthropic, which makes the Claude chatbot, and Google’s Gemini did not respond to requests for comment. OpenAI pointed to its posted election safeguards, which state the company continues to monitor bias and aims to keep ChatGPT’s responses politically neutral.
Concerns about AI
As Johnson refined his prompts, he outlined his general approach to voting, including how he weighs candidates aligned with the Make America Great Again movement. When he asked about immigration, ChatGPT separated topics such as border enforcement, asylum, and refugee resettlement.
It noted publicly available positions for Sen. Shelley Moore Capito, including support for stronger border enforcement, funding for ICE and CBP, the Laken Riley Act, and policies tied to the Trump administration, while saying there was less detailed immigration material available at the time for challenger Rachel Fetty Anderson and it would avoid assigning generic positions without evidence.
Johnson said he recognizes that chatbots can be sycophantic. “ChatGPT is just kind of a people pleaser,” he said, adding that it tends to frame a user’s preferences positively. Batista said simply knowing these systems are agreeable does not necessarily shield users from their influence.
Voters use chatbots in a variety of ways
In Madison, Wisconsin, longtime City Council staff alumna Lisa Veldran, 65, turned to Gemini for specific questions about the Democratic primary for governor after Lt. Gov. Sara Rodriguez exited the race. To compare Francesca Hong and David Crowley, she asked the chatbot to build a table of contrasts.
Veldran sees AI as a research aid, but emphasized, “You also have to keep in mind what’s running these chatbots and AI in general. Who’s paying to build these bots and what is their ultimate goal.”
In Exton, Pennsylvania, author Lakshmi Iyer, 50, who volunteers with They See Blue, an Indian American grassroots Democratic group focused on voter registration, said she did not use a chatbot to decide her own ballot. Instead, she used Claude to create Ballot Lookup, a website described as an independent, volunteer-built, nonpartisan tool that helps Pennsylvanians see who is on their ballot and where candidates stand.
To reduce bias, Iyer instructed the chatbot to source statements from candidates’ own speeches or official sites. She describes herself as a power user, but says “the human in the loop is absolutely essential.”
Some voters say shrinking local news coverage has left gaps, particularly for city and county contests. Tyler Black of Nashville said he does not rely on chatbots for congressional races, but finds them useful to learn about local candidates.
A software company vice president, Black knows chatbots can hallucinate, producing false or misleading information. He often checks one chatbot’s output against another. Batista said cross-model checks are better than asking the same system twice, though he has not seen firm evidence that this approach is consistently effective.
For those using AI chatbots for voting research, Batista recommends obscuring personal preferences in prompts and browsing in incognito mode so the system has less prior context about the user.
Back in West Virginia, Johnson is still working through his ballot with ChatGPT, with county commission contests next on his list. Overall, he said the technology has helped him prepare.
“I wouldn’t feel too bad if I had to do the election today,” he said.
How AI chatbots for voting can help, and what to watch for
- Useful for organizing races, building comparison tables, and surfacing candidate positions.
- Risk of subtle persuasion due to models mirroring user preferences.
- Possible hallucinations and opaque sourcing, requiring outside verification.
- Experts suggest hiding personal leanings in prompts and using private browsing.
Some of the immigration positions voters research with chatbots, including policies tied to the Trump administration, have also driven legal and political fights over border enforcement, such as when a judge halted certain work on the Texas border wall, as reported in Judge halts Big Bend border wall work across Texas sector.
Groups like They See Blue say volunteer-built tools and voter education remain important regardless of how AI evolves.
Voters and advocates watching for intimidation or misinformation at the polls have also followed federal statements closely, including when DHS said there were no plans for ICE operations at voting sites even as immigrant-rights groups mobilized.
